Knowledge Extraction

Overview

Knowledge Extraction component enables you to extract specific information and insights from processed knowledge bases. It provides powerful querying and extraction capabilities for targeted information retrieval.

What is Knowledge Extraction?

Knowledge Extraction is the process of retrieving specific information, patterns, and insights from structured knowledge bases using queries and extraction rules.

Key Features

Query-based Extraction: Extract information using natural language queries
Pattern Matching: Find patterns and relationships in knowledge
Insight Generation: Generate insights from extracted data
Multi-source Support: Extract from multiple knowledge sources
Customizable Rules: Define custom extraction rules

Basic Usage

import { KnowledgeExtraction } from 'gaik-toolkit';

// Create an extraction instance
const extractor = new KnowledgeExtraction();

// Extract information
const results = await extractor.extract(knowledge, query);

console.log(results);

API Reference

`KnowledgeExtraction`

Main class for extracting information from knowledge bases.

Constructor

const extractor = new KnowledgeExtraction(options);

Options:

queryMode (string): 'semantic' or 'keyword' (default: 'semantic')
maxResults (number): Maximum number of results (default: 10)
threshold (number): Relevance threshold 0-1 (default: 0.7)

Methods

`extract(knowledge: Knowledge, query: string): Promise<ExtractionResult>`

Extracts information based on a query.

Parameters:

knowledge: Knowledge object from KnowledgeCapture
query: Query string or extraction specification

Returns: Promise resolving to ExtractionResult object

Example:

const results = await extractor.extract(knowledge, 'Find all people mentioned');

`extractWithRules(knowledge: Knowledge, rules: ExtractionRules): Promise<ExtractionResult>`

Extracts information using custom rules.

Parameters:

knowledge: Knowledge object to extract from
rules: Custom extraction rules

Returns: Promise resolving to ExtractionResult object

`generateInsights(knowledge: Knowledge): Promise<Insights>`

Generates insights from knowledge base.

Parameters:

knowledge: Knowledge object to analyze

Returns: Promise resolving to Insights object

`ExtractionResult`

Result object containing extracted information.

Properties:

results (array): Extracted information items
confidence (number): Confidence score (0-1)
metadata (object): Extraction metadata
relatedItems (array): Related items found

`Insights`

Generated insights object.

Properties:

patterns (array): Identified patterns
trends (array): Detected trends
anomalies (array): Anomalies found
summary (string): Insight summary

Usage Examples

Basic Query Extraction

import { KnowledgeCapture, KnowledgeExtraction } from 'gaik-toolkit';

const capture = new KnowledgeCapture();
const extractor = new KnowledgeExtraction();

// Capture knowledge
const knowledge = await capture.captureFromText(text);

// Extract specific information
const people = await extractor.extract(knowledge, 'List all people mentioned');
const dates = await extractor.extract(knowledge, 'Find important dates');

Semantic Search

const extractor = new KnowledgeExtraction({
  queryMode: 'semantic',
  maxResults: 5,
  threshold: 0.8
});

const results = await extractor.extract(
  knowledge,
  'What are the main topics discussed?'
);

results.results.forEach(item => {
  console.log(`${item.content} (confidence: ${item.confidence})`);
});

Custom Extraction Rules

const rules = {
  entities: ['person', 'organization'],
  relationships: ['works_for', 'located_in'],
  filters: {
    minConfidence: 0.8,
    dateRange: { start: '2023-01-01', end: '2024-12-31' }
  }
};

const results = await extractor.extractWithRules(knowledge, rules);

Insight Generation

const insights = await extractor.generateInsights(knowledge);

console.log('Key Patterns:', insights.patterns);
console.log('Trends:', insights.trends);
console.log('Summary:', insights.summary);

Advanced Features

Multi-source Extraction

// Extract from multiple knowledge bases
const knowledgeBases = [knowledge1, knowledge2, knowledge3];

const allResults = await Promise.all(
  knowledgeBases.map(kb => extractor.extract(kb, query))
);

// Merge results
const merged = extractor.mergeResults(allResults);

Batch Queries

const queries = [
  'Find all people',
  'List organizations',
  'Extract dates'
];

const batchResults = await extractor.batchExtract(knowledge, queries);

Result Filtering and Ranking

const results = await extractor.extract(knowledge, query);

// Filter by confidence
const highConfidence = results.results.filter(r => r.confidence > 0.9);

// Sort by relevance
const sorted = results.results.sort((a, b) => b.confidence - a.confidence);

Complete Workflow Example

import { Parser, KnowledgeCapture, KnowledgeExtraction } from 'gaik-toolkit';

async function processDocument(filePath, query) {
  // Parse document
  const parser = new Parser();
  const document = await parser.parse(filePath);

  // Capture knowledge
  const capture = new KnowledgeCapture();
  const knowledge = await capture.captureFromDocument(document);

  // Extract information
  const extractor = new KnowledgeExtraction();
  const results = await extractor.extract(knowledge, query);

  // Generate insights
  const insights = await extractor.generateInsights(knowledge);

  return {
    extractedInfo: results,
    insights: insights
  };
}

// Use it
const result = await processDocument('document.pdf', 'Summarize key findings');
console.log(result);

Error Handling

try {
  const results = await extractor.extract(knowledge, query);

  if (results.results.length === 0) {
    console.log('No results found for query');
  }
} catch (error) {
  if (error.code === 'INVALID_QUERY') {
    console.error('Query format is invalid');
  } else if (error.code === 'EXTRACTION_FAILED') {
    console.error('Extraction process failed:', error.message);
  } else {
    console.error('Unexpected error:', error);
  }
}

Overview

What is Knowledge Extraction?

Key Features

Basic Usage

API Reference

`KnowledgeExtraction`

Constructor

Methods

`extract(knowledge: Knowledge, query: string): Promise<ExtractionResult>`

`extractWithRules(knowledge: Knowledge, rules: ExtractionRules): Promise<ExtractionResult>`

`generateInsights(knowledge: Knowledge): Promise<Insights>`

`ExtractionResult`

`Insights`

Usage Examples

Basic Query Extraction

Semantic Search

Custom Extraction Rules

Insight Generation

Advanced Features

Multi-source Extraction

Batch Queries

Result Filtering and Ranking

Complete Workflow Example

Error Handling

Next Steps

On this page